Bootstrapping The Autoregressivedistributed Lag Test For Cointegration
Loading...
Date
2016-01
Authors
Sam, Chung Yan
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
The objective of this thesis is to examine the performances of a cointegration test:
Autoregressive Distributed Lag (ARDL) bounds test approach developed by Pesaran et
al. (2001). This approach gained popularity and is widely used for over two decades due
to its advantages of super consistent estimation and dealing with mixed integration order
regressors. Nevertheless, the ARDL bounds test is often applied in environments that are
inconsistent with the assumptions underlying that framework. This approach assumes that
there is no feedback at level from the dependent variable to the regressors. That is, all
variables except one must be weakly exogenous. Estimation involving several plausibly
endogenous variables as regressors will give biased and misleading results. However,
through simulation evidence our results show that the performance of the bounds test
approach is not affected by this endogeneity problem. In this thesis, we propose a new
ARDL cointegration test that relies on the bootstrap procedure. It is shown that by
introducing a proper bootstrap procedures, some weaknesses underlying the approach are
improved based on size and power properties. In addition, it eliminates the possibility of
inconclusive inferences from bounds testing. Besides that, inference based solely on the
significance of F test and single t test is insufficient to avoid degenerate case. Conducting
an additional testing on the lagged independent variable comes from the proposed method
to provide a better insight in concluding the status of cointegration. The empirical
relevance of the bootstrap ARDL test is demonstrated by an estimation of savinginvestment
correlations.
Description
Keywords
Autoregressive Distributed Lag (ARDL)